Imperfectly Rational Models Of Addiction
Several extensions followed the Becker and Murphy (1988) model to address the restrictive assumptions of perfect rationality. Two assumptions that have received attention are the assumption of perfect information and foresight and the assumption of exponential discounting. Models that address these concerns otherwise follow a common strategy to Becker and Murphy (1988); people continue to make decisions that they believe – at the time the decision is made – are in their best interest.
The perfectly rational consumer correctly predicts the effect that consumption of an addictive good will have on their behavior in the future. However, this assumption is contrary to evidence that suggests people are very poor judges of their future preferences and tastes: People tend to bias estimates of their future tastes toward being like their current tastes (Loewenstein et al., 2003). This utility projection bias is particularly troublesome when nonaddicted people need to make judgments about the impact that consumption of an addictive good will have on their future preferences . Badger et al. (2007) show that even seasoned heroin addicts underestimate the influence of their addiction on behavior. To address this issue, Orphanides and Zervos (1995) extend the rational addiction model to allow people to be uncertain about how addictive they will find a good or activity. People update their beliefs about the addictiveness of the good by observing the actions of those around them and through their own experimentation. People try their first cigarette, for example, without knowing how addictive they will find smoking. In this model, addicts may regret their past choices even though they make the best choices they can with the information available at the time. Some people will underestimate their potential for addiction and regret having become an addict.
For most addictive goods, such as cigarettes or narcotics, consumption leads to an immediate benefit while the cost, such as poor health, is realized in the future. For this reason the manner in which people discount the future has important consequences for the rational addiction model. Dynamic rationality implies that people discount exponentially and consistently. That is, a predetermined and constant rate of discount is applied to every period in the future. Models of hyperbolic discounting instead assume that people have a present bias, applying a larger discount rate to events far in the future than events that are to occur sooner. A number of controlled experiments find that hyperbolic discounting is a more accurate depiction of behavior than exponential discounting (for a review of the evidence on hyperbolic discounting see Frederick et al. (2002)). If this type of discounting accurately reflects the decision process, then people will underweigh the future costs of their actions at the time decisions are made. Gruber and Koszegi (2001) extend the perfectly rational model to include hyperbolic discounting. This model yields dramatically different normative implications than the Becker and Murphy (1988) framework, as there is an ‘internality’ – one’s smoking today harms one’s future self, and one’s present self and one’s future self are, in effect, in conflict.
In the canonical rational addiction model, and some extensions thereof, people make lifetime consumption plans and adjust them as new information is revealed. A different approach to modeling the behavior of a rational addict is taken by Gul and Pesendorfer (2007) who consider people who make a consumption plan, but need to exert costly self-control to see it through. Consider a rational alcoholic who determines an optimal consumption plan of four drinks per day. According to the Becker and Murphy framework, absent any changes in information, the rational alcoholic will see this plan through, consuming four and only four drinks daily. However, such a plan requires self-control. The temptation of having extra alcohol in the house may cause the alcoholic to deviate from the four drink per day plan, and instead have five or six drinks. This deviation from the plan in the current period makes self-control in future periods even more difficult. In this framework, an addictive good is harmful if people experience an ever-widening gap between their planned optimal consumption and their actual consumption. Like the Becker and Murphy model, addicts in this model respond to anticipated future price increases by decreasing current consumption patterns, and may exhibit binging and abstinence cycles. However, unlike the Becker and Murphy model, this model can explain the use of short-term commitment devices, such as rehabilitation centers, by addicts.
Irrational Models Of Addiction
Most researchers outside the field of economics do not think about addiction in a rational decision framework. This largely follows from an empirical anomaly: Addicts commonly express a strong desire to reduce or stop their consumption of addictive goods but fail to follow through. The ability of addiction to override rationality is captured in a statement made by David Kessler, former commissioner of the Food and Drug Administration: ‘‘Once they have started smoking regularly, most smokers are in effect deprived of the choice to stop smoking’’ (statement to the House Subcommittee on Health and the Environment 25 March 1994).
The economist views a consumer as irrational if decision making ignores relevant information and incentives. For example, models of myopic decision making can be thought of as irrational; people do not consider how current decisions will impact future outcomes. It has been argued that addiction leads to a failure in the processing of information, and therefore causes the addict to deviate from rational decision making. Clinical evidence suggests that addicts exhibit a bias in their mental accounting, placing too little weight on the negative consequences of their behavior (see Tomer (2001) for a discussion). Further, the observed procrastination of addicts, wishing to quit but continually putting off action, suggests that rational behavior does not fully capture addictive behavior.
Even if the consumption of addictive substances is irrational, surely addicts are rational in some facets of their lives. Bernheim and Rangel (2004) model a ‘cue-triggered’ decision process built on three premises. First, the consumption of addictive goods by an addict is often a mistake. Second, increased consumption of addictive goods makes addicts more sensitive to random environmental cues that trigger mistaken consumption. Third, addicts understand the cue-triggered process and take steps to manage their susceptibility. The cue-trigger model draws on neurological evidence that addictive substances interfere with the operation of pleasure and reward processes in the brain. In this model, people face a dynamic decision process in which environmental cues trigger a ‘hot’ decision-making mode during which the substance is consumed regardless of relevant incentives and information. When operating in a ‘cold’ mode people fully consider the current and future consequences of their actions, including how decisions influence the likelihood of being cued into a hot mode. For example, if stressful circumstances exacerbate the cravings associated with cigarette addiction, then a person trying to quit smoking will likely take steps to avoid stressful circumstances. Addicts in this model are aware of their propensity to make consumption mistakes and will take steps to precommit to future consumption and mitigate cues.